How to derive this rank equality for block matrices?

183 Views Asked by At

Consider the following:

$$\begin{bmatrix}X & I \\ I & Y\end{bmatrix}=\begin{bmatrix}I & 0 \\ Y^{-1} & I\end{bmatrix}^T\begin{bmatrix}X-Y^{-1} & 0 \\ 0 & Y\end{bmatrix}\begin{bmatrix}I & 0 \\ Y^{-1} & I\end{bmatrix}$$

where

  1. $X,Y\in \mathbb{R}^{n\times n}\succ0$ (positive definite)
  2. $I\in \mathbb{R^{n\times n}}$ is the identity matrix.
  3. Suppose we know $$\begin{bmatrix}X & I \\ I & Y\end{bmatrix}\succeq 0$$

How to show the following: $$\textbf{rank}\begin{bmatrix}X & I \\ I & Y\end{bmatrix}=n+\textbf{rank}(X-Y^{-1})=n+\textbf{rank}(XY-I)$$

I cannot find a way to figure out why this is true.

1

There are 1 best solutions below

0
On BEST ANSWER

Multiplying a matrix $A$ from the left or from the right by an invertible matrix doesn't change its rank. The matrix

$$ \begin{pmatrix} I & 0 \\ Y^{-1} & I \end{pmatrix} $$

is invertible (being an lower block triangular with invertible blocks on the diagonal) and so is its transpose and hence

$$ \operatorname{rank} \begin{pmatrix} X & I \\ I & Y \end{pmatrix} = \operatorname{rank} \begin{pmatrix} X - Y^{-1} & 0 \\ 0 & Y \end{pmatrix}. $$

The rank of a block diagonal matrix is the sum of ranks of the blocks on the diagonal and so

$$ \operatorname{rank} \begin{pmatrix} X - Y^{-1} & 0 \\ 0 & Y \end{pmatrix} = \operatorname{rank} (X - Y^{-1}) + \operatorname{rank} Y. $$

Since $Y > 0$, we have $\operatorname{rank} Y = n$ giving the first equality. Since $$\operatorname{rank} (X - Y^{-1}) = \operatorname{rank} ((X - Y^{-1})Y) = \operatorname{rank}( XY - I) $$ you get the second equality.